Journal of Algebra 324 (2010) 3035–3043

Contents lists available at ScienceDirect

Journal of Algebra
www.elsevier.com/locate/jalgebra

Invariant ideals of abelian group algebras under the torus

action of a field, I ✩
D.S. Passman
Department of Mathematics, University of Wisconsin–Madison, Madison, WI 53706, USA

a r t i c l e i n f o a b s t r a c t

Article history: Let V = V 1 ⊕ V 2 be a finite-dimensional vector space over

Received 26 January 2010 an infinite locally-finite field F . Then V admits the torus action
Available online 4 February 2010 of G = F • by defining ( v 1 ⊕ v 2 ) g = v 1 g −1 ⊕ v 2 g. If K is a field of
Communicated by Nicolás Andruskiewitsch
characteristic different from that of F , then G acts on the group
and Robert Guralnick
algebra K [ V ] and it is an interesting problem to determine all
Happy Birthday, Susan G-stable ideals of this algebra. In this paper, we consider the
special case when V 1 and V 2 are both 1-dimensional and we show
Keywords: that there are just four G-stable proper ideals of K [ V ].
Group algebra © 2010 Elsevier Inc. All rights reserved.
Group action
Torus action
Invariant ideal

1. Introduction

Let K be a field, let V be an abelian group, and let G be a group of automorphisms of V . Then G
acts on the group algebra K [ V ] and it is an interesting, and surprisingly difficult, problem to describe
the G-stable ideals of K [ V ]. The motivation for this actually comes from the study of the lattice of
ideals in group algebras of certain infinite locally finite groups. This can be seen, for example, in the
survey [6] or in the introduction to paper [7], but we will not expand upon this theme here. A natural
special case of the problem occurs when V is a vector space over an infinite field F and when G = F •
acts on V by scalar multiplication. This turns out to have a rather beautiful solution [1,7,4], especially
when V is finite-dimensional. Indeed, one can even allow F to be a division algebra.
The next case of interest surely arises by introducing inverses from F . Specifically, let V 1 and V 2 be
two vector spaces over F , form V = V 1 ⊕ V 2 , and let G = F • act on V by ( v 1 ⊕ v 2 ) g = v 1 g −1 ⊕ v 2 g.
We call this the torus action of F by analogy to the way the torus in SL2 ( F ) acts on F 2 . The goal
of this paper is to describe the G-stable ideals of K [ V ] when dim F V 1 = dim F V 2 = 1 and when F is

✩
Research supported in part by an NSA grant.
E-mail address: passman@math.wisc.edu.

0021-8693/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jalgebra.2010.01.023
3036 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043

an infinite locally finite field. As we will see, there are just four G-stable proper ideals in this situation.
The argument here is tricky, but not difficult, and is hopefully a first step towards a solution of the
general problem.
Let V be an abelian group, viewed multiplicatively, and let K [ V ] denote its group algebra over
the field K . If A is a subgroup of V , then there exists a natural epimorphism K [ V ] → K [ V / A ] and
we let ω( A ; V ) = ω K ( A ; V ), the augmentation ideal of A in V , denote its kernel. Thus, ω( A ; V ) is the
K -linear span of all elements of the form (1 − a) v with a ∈ A and v ∈ V , and clearly

  
A = v ∈ V  1 − v ∈ ω( A ; V ) .

Observe that if A and B are subgroups of V and if C =  A , B  is the group they generate,
then ω( A ; V ) + ω( B ; V ) = ω(C ; V ). Indeed, if I denotes the ideal ω( A ; V ) + ω( B ; V ), then surely
I ⊆ ω(C ; V ). On the other hand, both A and B are contained in the kernel of the homomorphism
K [ V ] → K [ V ]/ I restricted to the group V , and hence C is also contained in this kernel. Now, if G is a
group that acts as automorphisms on V , then G also acts on K [ V ], and it is clear that A is a G-stable
subgroup of V if and only if ω( A ; V ) is a G-stable ideal of K [ V ].
We now return to additive notation for V . The main result of this paper is

Theorem 1.1. Let F be an infinite locally-finite field, let V 1 and V 2 be 1-dimensional F -vector spaces, and set
V = V 1 ⊕ V 2 . If G = F • , then we can let G act as the torus on V , and hence G acts on the group algebra K [ V ].
Suppose, in addition, that char K = char F . Then the nontrivial G-stable ideals of K [ V ] are precisely ω( V ; V ),
ω( V 1 ; V ), ω( V 2 ; V ) and ω( V 1 ; V ) ∩ ω( V 2 ; V ).

Note that, if V is a torsion abelian group having no elements of order equal to the characteristic
of K , then K [ V ] is a commutative von Neumann regular algebra (see [5, Theorem 1.1.5]). It follows
that if I , J
K [ V ], then I ∩ J = I J . In particular, finite products and finite intersections of ideals
coincide here. Furthermore, every ideal of K [ V ] is semiprime.
We close this section by describing the G-stable subgroups of V = V 1 ⊕ V 2 when V 1 and V 2 are
arbitrary F -vector spaces. The argument is slightly simpler in the case of locally-finite fields, but we
prove the result in full generality.

Lemma 1.2. Let F be a field with | F |  5. If V = V 1 ⊕ V 2 is an F -vector space admitting the torus action of
G = F • , then the G-stable subgroups of V are precisely those of the form A ⊕ B, where A is an F -subspace of
V 1 and B is an F -subspace of V 2 .

Proof. Obviously, all such A ⊕ B are G-stable subgroups of V . We consider the converse. To this
end, let W 1 = W 2 = F be the 1-dimensional G-modules given by w 1 = w 1 g −1 and w 2 = w 2 g for
g g
•
all w 1 ∈ W 1 , w 2 ∈ W 2 and g ∈ G = F . Then W 1 and W 2 are both irreducible G-modules since
any G-submodule of W i is closed under addition and scalar multiplication by F . We claim now that
W 1 and W 2 are not G-isomorphic. Indeed, if such an isomorphism θ : W 1 → W 2 exists, then for all
−1 −1
nonzero elements f ∈ F , we have θ( f ) = θ(1· f ) = θ(1 f ) = θ(1) f = θ(1) f −1 . Thus, if α , β and
α + β are nonzero elements of F , then θ(α ) = θ(1)α , θ(β) = θ(1)β −1 and
−1

 
θ(1) α −1 + β −1 = θ(α ) + θ(β) = θ(α + β) = θ(1)(α + β)−1 .

Hence α −1 + β −1 = (α + β)−1 . In particular, if x = 0, 1 is in F , then setting α = x and β = 1 − x, we

obtain x−1 + (1 − x)−1 = 1 and hence x satisfies x2 − x + 1 = 0. There are, of course, at most two
solutions to the latter equation, so | F |  4 contrary to our hypothesis.
Returning to the vector space V = V 1 ⊕ V 2 , we see that V 1 is a direct sum of G-submodules
isomorphic to W 1 , and V 2 is a direct sum of G-submodules isomorphic to W 2 . Thus V is a completely
reducible G-module and hence so is any submodule. Indeed, any submodule is a direct sum of copies
 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043 3037

of W 1 and of W 2 . But W 1  W 2 , so any copy of W 1 in V is contained in V 1 , and any copy of W 2 in

V is contained in V 2 . Thus, any G-submodule of V is of the form A ⊕ B, as required. 2

We actually showed above that if W 1 ∼

= W 2 , then the inverse map in F extends to a field auto-
morphism, and this does occur when F = GF(2), GF(3), or GF(4).

2. Finite fields

In this section we obtain a few combinatorial results on finite fields. We start with a corollary to
a simple special case of a result of [3]. We include the quick proof for the convenience of the reader.

Lemma 2.1. Let E be a finite field, let V = 0 be an additive subgroup of E + , and let 0 = s ∈ E. If sx−1 ∈ V for
all 0 = x ∈ V , then V = Lt where L is a subfield of E and t 2 ∈ Ls.

Proof. Let x, y ∈ V with xy = 0 or s. Then 0 = (xy − s)/ y = x − (s/ y ) ∈ V , and hence sy /(xy − s) ∈ V .
Thus xy 2 /(xy − s) = y + sy /(xy − s) ∈ V and, by taking inverses and multiplying by s, we have
s(xy − s)/xy 2 ∈ V . It follows that s2 /xy 2 = (s/ y ) − s(xy − s)/xy 2 ∈ V and therefore xy 2 /s =
s(xy 2 /s2 ) ∈ V . Of course, this inclusion is satisfied when xy = 0 or s, so we see that xy 2 /s ∈ V for all
x, y ∈ V .
Now let L = {r ∈ E | V r ⊆ V }. Then L is a subring of E and hence also a subfield, so V is an
L-vector space. Notice that xy 2 /s ∈ V for all x, y ∈ V , so y 2 /s ∈ L and thus, since 0 ∈ V , we have
| L | > | V |/2. In particular, if dim L V  2, then | L | > | V |/2  | L |2 /2, so 2 > | L |, a contradiction. Thus
dim L V = 1 and V = Lt for some 0 = t ∈ E. Finally, since s/t ∈ V = Lt, we have t 2 ∈ Ls. 2

Next, we extend the above argument to prove

Lemma 2.2. Let F ⊆ E be finite fields with | E : F |  3, let 0 = s ∈ E, and let λ: E → F be an F -linear functional
with λ(1) = 1. Define the subset H of F by

   
H = λ sx−1  x ∈ E , λ(x) = 1 .

Then 0 ∈ H and there exists a fixed 0 = b ∈ F , such that every element a ∈ F satisfies a polynomial equation
of the form

a2 τ b − ab + σ = 0
√
for suitable τ , σ ∈ H depending upon a. In particular, | H |  | F |/2.

Proof. Let | F | = q and | E | = qn with n = | E : F |. If V = ker λ, then V is an F -subspace of E of dimen-

sion n − 1. Furthermore,

  
V + 1 = r ∈ E  λ(r ) = 1 .

Suppose sv −1 ∈ V for all 0 = v ∈ V . Then, by the preceding lemma, V = Lt for some proper subfield
L of E and some 0 = t ∈ E. Indeed, since F V ⊆ V , we have F ⊆ L and hence

n − 1 = dim F V = dim F L = | L : F |  n/2,

contradicting the assumption that n  3.

Thus there exists an element 0 = v ∈ V with s/ v ∈ / V . Then 0 = d = λ(s/ v ) ∈ F , and if we
set y = s/ vd, then we have λ( y ) = 1 and s/ y = vd ∈ V . In particular, 0 = λ(sy −1 ) ∈ H . Note that
3038 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043

0∈ / V + 1, so the map V + 1 → E given by x → s2 /xy 2 has image of size | V | and this image does
not contain 0. Thus, this image cannot be contained entirely within V . We now fix x ∈ V + 1 with
s2 /xy 2 ∈
/ V , and we set 0 = b = λ(s2 /xy 2 ). To reiterate, we have fixed x, y ∈ V + 1 with s/ y ∈ V and
with 0 = b = λ(s2 /xy 2 ) ∈ F .
Let a ∈ F be arbitrary. Then as/ y ∈ V , so

xy + as as
=x+ ∈ V +1
y y

and we set τ = τa = λ(sy /(xy + as)) ∈ F , so that τ ∈ H . For convenience, define

xy 2 asy
z= =y−
xy + as xy + as

and let c = ca = λ( z). Since λ( y ) = 1 and λ(asy /(xy + as)) = aτ , we have c = 1 − aτ . If c = 0, then a
clearly satisfies

a2 τ b − ab + σ = σ − cab = 0

with 0 = σ ∈ H .
Now suppose that c = 0. Thus z/c ∈ V + 1 and we let σ = σa = λ(cs/z) so that σ ∈ H . Note that

cas2 cs(xy + as) cs cs cs

= − = −
xy 2 xy 2 y z y

so cab = caλ(s2 /xy 2 ) = λ(cas2 /xy 2 ) = σ , since λ(cs/ z) = σ and λ(cs/ y ) = 0. Also using c = 1 − aτ ,
we have (1 − aτ )ab = cab = σ and thus a does indeed satisfy the polynomial equation

a2 τ b − ab + σ = 0

determined by σ and τ .
Write h = | H | and note that b = 0. Then each pair (τ , σ ) ∈ H × H determines a unique nonzero
polynomial of degree  2 given by

ζ 2 τ b − ζ b + σ ∈ F [ζ ]

and hence we have h2 of these. Furthermore, each of these has at most two roots in F , so we obtain
√ shown above that each a ∈ F is a root
at most 2h2 roots in this manner. On the other hand, we have
of at least one of these polynomials. Thus, 2h2  q and h  q/2, as required. 2

We can sharpen the latter inequality a bit by considering separately those polynomials with either
τ = 0 or σ = 0 since they have at most one nonzero root.
We also need some results on linear functionals. Let V be a vector space over the finite field F
of size | F | = q and let M = { v 1 , v 2 , . . . , v m } be a linearly independent subset of V of size m. If 
V
denotes the dual space of V , define

  
V  λ( M ) = F
On( M ) = λ ∈ 

and

  
V  λ( M ) < F .
Non( M ) = λ ∈ 
 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043 3039

Thus On( M ) is the set of linear functionals that map M onto F , and Non( M ) is its complement in V.
Obviously, On( M ) = ∅ if and only if m < q. In the following, we let ln denote the natural logarithm.

Lemma 2.3. Let V be a vector space over F with | F | = q and dim F V = n. In addition, let M be a linearly
independent subset of V of size m and suppose that m  2q ln q. Then |Non( M )| < qn−1 and |On( M )| >
qn − qn−1 . In particular, the latter two inequalities hold when m  q2 .

Proof. Extend M to a basis of V by adding the vectors w 1 , w 2 , . . . , w k with m + k = n. We obtain a

quick upper bound for |Non( M )| by using the first term of inclusion–exclusion. Specifically, note that
there are q subsets F 1 , F 2 , . . . , F q of F of size q − 1, and Non( M ) is the union over i of all linear
functionals λ with λ( M ) ⊆ F i . Now for each i and j, there are q − 1 choices for λ( v j ) ∈ F i and q
choices for λ( w j ) ∈ F . Thus the number of functionals with λ( M ) ⊆ F i is precisely (q − 1)m qk , and
hence |Non( M )|  q·(q − 1)m qk .
For this result we want |Non( M )| < qn−1 = qm+k−1 , and so it suffices to have q·(q − 1)m qk < qm+k−1
or (q − 1)m < qm−2 . Taking logarithms, this is equivalent to m ln(q − 1) < (m − 2) ln q or 2 ln q <
m(ln q − ln(q − 1)). Since ln q − ln(q − 1) > 1/q, it therefore suffices to have m/q  2 ln q or m  2q ln q.
Since On( M ) is the complement of Non( M ) in  V , we see that m  2q ln q implies that |On( M )| >
qn − qn−1 . Finally, q > 2 ln q, so m  q2 implies m  2q ln q. 2

Of course, |On( M )| can be described precisely using full inclusion–exclusion. Its m-part, as given
in the above proof, can also be written as q! S (m, q), where S (m, q) denotes the Stirling number of the
second kind (see for example [2, pp. 287 and 317]). We close with a well-known observation.

Lemma 2.4. Let E ⊇ F be fields with | E : F | < ∞ and let λ: E → F be a nonzero F -linear functional. Then
every F -linear functional from E to F is uniquely of the form λa for a ∈ E, where λa (x) = λ(ax).

Proof. The map a → λa is easily seen to be an F -linear transformation from E to 

E, and since E is a
field, this map is one-to-one. By dimension considerations, the map is therefore also onto. 2

3. G -stable ideals

We now begin our work on Theorem 1.1. In the following E, F and L will denote fields of charac-
teristic p > 0. For the most part, they will be finite or at least locally finite. In particular, they will be
subfields of a fixed algebraic closure of GF( p ). Furthermore, any vector space over any of these fields
is additively an elementary abelian p-group. In addition, we let K be a field of characteristic different
from p, and we assume until further notice that K is algebraically closed or at least that it contains a
primitive pth root of unity ε . We fix the prime p and the field K throughout. The following facts are
standard.

Lemma 3.1. Let V be a finite elementary abelian p-group and let G be a group of automorphisms of V .

(i) The group algebra K [ V ] is semisimple. Indeed, it is a direct sum of | V | copies of K and every ideal is
uniquely an intersection of maximal ideals.
(ii) The maximal ideals of K [ V ] are in one-to-one correspondence with the linear characters χ : V → K • . To
be precise, the ideal corresponding to χ is the kernel of the natural algebra extension χ : K [ V ] → K .
−1
(iii) G permutes the linear characters of V by defining χ g (x) = χ (x g ) for all g ∈ G and x ∈ V . This action
corresponds to the permutation action of G on the maximal ideals of K [ V ].
(iv) Every G-stable ideal of K [ V ] is uniquely an intersection of the maximal G-stable ideals of K [ V ]. The latter
are precisely the intersections of G-orbits of maximal ideals of K [ V ].
−1
(v) χ g = χ if and only if x−1 x g ∈ ker χ for all x ∈ V .

Now suppose F and V = V 1 ⊕ V 2 are both finite. If g ∈ G = F • fixes a character χ of V , then

by (v) above, we have ker χ ⊇ V 1 ( g − 1) + V 2 ( g −1 − 1), in additive notation. In particular, if g = 1,
3040 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043

then ker χ ⊇ V and χ is the trivial (i.e. principal) character of V . In other words, G permutes all the
nontrivial characters of V in orbits of full size |G |.

Lemma 3.2. Let F be a finite field with | F | = q and let G = F • act as the torus on V = V 1 ⊕ V 2 , where both
V 1 and V 2 are 1-dimensional.

(i) K [ V 1 ] has precisely two maximal G-stable ideals, namely ω( V 1 ; V 1 ) and one other which we denote
by J 1 . They satisfy ω( V 1 ; V 1 ) ∩ J 1 = 0.
(ii) K [ V 2 ] has precisely two maximal G-stable ideals, namely ω( V 2 ; V 2 ) and one other which we denote
by J 2 . They satisfy ω( V 2 ; V 2 ) ∩ J 2 = 0.
(iii) There are q + 2 maximal G-stable ideals of K [ V ]. One is ω( V ; V ), and two others J 1 and J 2 satisfy
J 1 ∩ ω( V ; V ) = ω( V 1 ; V ) and J 2 ∩ ω( V ; V ) = ω( V 2 ; V ). For convenience, J 1 and J 2 are said to be
quasi-augmentation ideals, while the remaining q − 1 maximal G-stable ideals different from ω( V ; V )
are said to be standard.

Proof. We know that G permutes the nontrivial characters of K [ V 1 ], K [ V 2 ] and K [ V ] in orbits of

full size |G | = q − 1. In particular, since | V 1 | = q, it follows that there are just two maximal G-stable
ideals of K [ V 1 ], namely ω( V 1 ; V 1 ) and J 1 . Clearly, ω( V 1 ; V 1 ) ∩ J 1 = 0, so (i) is proved, and (ii) follows
similarly.
For (iii), we have | V | = q2 , so there are 1 + (q2 − 1)/(q − 1) = q + 2 G-orbits on the characters
of V and hence on the maximal ideals of K [ V ]. The trivial character of course corresponds to the
augmentation ideal ω( V ; V ). Since ω( V 1 ; V ) is a G-stable ideal of codimension q, we see that there is
a maximal G-stable ideal J 1 with J 1 ∩ ω( V ; V ) = ω( V 1 ; V ). Similarly, there exists a maximal G-stable
ideal J 2 with J 2 ∩ ω( V ; V ) = ω( V 2 ; V ). We have accounted for 3 of the maximal G-stable ideals of
K [ V ], so there are (q + 2) − 3 = q − 1 remaining ideals which we consider to be standard. 2

While the above gives us a quick count on the G-stable ideals of K [ V ], it does not really give us a
good description of them. So we take a closer look at the action of G = F • on V = F 2 and on K [ V ].

Example 3.3. Let F be a finite field and let V = F 2 . Then the torus action of F • on F 2 is given by
(a, b) f = ( f −1 a, f b) for all a, b ∈ F and f ∈ F • . Of course, F • also acts on the group algebra K [ F 2 ]
and our goal here is to describe the maximal F • -stable ideals of K [ F 2 ]. As usual, we assume that
char F = p > 0, char K = p and that K contains ε , a primitive pth root of unity.
Let GF( p ) denote the prime subfield of F and let μ: F → GF( p ) be a nonzero GF( p )-linear func-
tional. Then, by Lemma 2.4, all linear functionals from F to GF( p ) are of the form μa : F → GF( p )
where a ∈ F and μa (x) = μ(ax). Hence all characters χ : F → K • are given by χa (x) = ε μa (x) = ε μ(ax) .
Furthermore, since the characters from F 2 to K • are necessarily products, they are all of the form

χa,b (x, y ) = εμ(ax)+μ(by) = εμ(ax+by) .

These in turn extend to K -algebra homomorphisms χa,b : K [ F 2 ] → K and their kernels I a,b = ker χa,b
are precisely the set of maximal ideals of K [ F 2 ].
Now F • permutes these characters by

 −1   
χag,b (x, y ) = χa,b (x, y ) g = χa,b gx, g −1 y
−1 y )
= ε μ(agx+bg = χag ,bg −1 (x, y )

for all g ∈ F • , and hence we have

g
I a,b = I ag ,bg −1 .
 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043 3041

The F • -orbits of these ideals are now easily seen to be

O0,0 = { I 0,0 },
O0,∗ = { I 0,b | 0 = b ∈ F },
O∗,0 = { I a,0 | 0 = a ∈ F },
Od = { I a,b | a, b ∈ F , ab = d} for all 0 = d ∈ F .

Furthermore, the maximal F • -stable ideals of K [ F 2 ] are precisely the intersections of the kernels I a,b
over all members of an orbit. Thus, we can denote the maximal F • -stable ideals of K [ F 2 ] by I 0,0 , I 0,∗ ,
I ∗,0 , and I d for all 0 = d ∈ F , where the subscripts of course correspond to the orbit notation.
It is easy to see that I 0,0 = ω( F 2 ; F 2 ) is the augmentation ideal of K [ F 2 ]. Furthermore I 0,0 ∩ I 0,∗ =
ω( F ⊕ 0; F 2 ) and I 0,0 ∩ I ∗,0 = ω(0 ⊕ F ; F 2 ), where F ⊕ 0 and 0 ⊕ F are the obvious F • -stable subgroups
of F 2 . In other words, I 0,∗ and I ∗,0 are the quasi-augmentation maximal F • -stable ideals of K [ F 2 ],
while the various I d are the standard maximal F • -stable ideals. Of course, the intersection of all these
maximal F • -stable ideals in 0.

Our goal now is to show that standard ideals do not appear in certain situations.

Lemma 3.4. Let E ⊇ L ⊇ F be finite fields with | F | = q. Suppose that | E : L |  3, | L : F |  2q2 + 1, and let E 2
admit the torus action of E • . If I is a standard maximal E • -stable ideal of K [ E 2 ], then I ∩ K [ F 2 ] = 0.

Proof. We use the description and notation for K [ F 2 ] as given in the preceding example. Of course,
that example applies equally well to K [ E 2 ] and K [ L 2 ] provided we have appropriate functionals to
the prime subfield GF( p ). For this, we choose λ: E → L, an L-linear functional with λ(1) = 1, and we
choose η: L → F , an F -linear functional with η(1) = 1. Then the composite functional η̄ = μη maps
L to GF( p ), while λ̄ = μηλ = η̄λ maps E to GF( p ).
Now we are given a standard maximal E • -stable ideal I of K [ E 2 ]. Using the functional λ̄, I is then
equal to I s for some 0 = s ∈ E. In other words, I is the intersection of the kernels of the algebra
homomorphisms K [ E 2 ] → K associated with the characters

χa,b (x, y ) = ελ̄(ax+by) = εη̄λ(ax+by)

for all a, b ∈ E with ab = s. Alternately, we can write a = sz−1 and b = z for all 0 = z ∈ E.
Since I a,b is the kernel of algebra homomorphism χa,b : K [ E 2 ] → K , it follows that I a,b ∩ K [ L 2 ] is
the kernel of the restriction of χa,b to K [ L 2 ]. Furthermore, since λ: E → L is an L-linear functional,
we see that λ(ax + by ) = λ(a)x + λ(b) y for all x, y ∈ L. Thus, with a = sz−1 and b = z, we see that the
restriction of the character χa,b to L 2 is given by

−1 )x+λ( z) y )
χ̃ (x, y ) = εη̄(λ(sz .

In particular, if we consider only those z with λ( z) = 1, then the various χ̃ that are obtained in this
manner correspond to L • -orbits with product given by λ(sz−1 ). Since | E : L |  3, Lemma 2.2 tells us
that

   
H = λ sz−1  z ∈ E , λ( z) = 1
√
satisfies 0 ∈ H and | H |  | L |/2.
To reiterate, if we use χ̃ to denote the characters of L 2 , then we have shown that the L • -stable
ideal I ∩ K [ L 2 ] is contained in the intersection of the kernel ideals corresponding to the L • -orbits of
the characters
3042 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043

χ̃h,1 = εη̄(hx+ y)

for all h ∈ H . Thus we have obtained a reasonably large number of L • -orbits in this restriction, but
not enough to guarantee that I ∩ K [ L 2 ] = 0.
So, we require a second pass, this time from L to F√. If m = q2 , then by assumption, n = | L : F | 
2m + 1. Thus | L |/2  q2m+1 /2  q2m and hence | H |  | L |/2  qm . It follows that the F -linear span
of the elements of H has F -dimension at least m, and hence we can choose h1 , h2 , . . . , hm ∈ H so that
M = {h1 , h2 , . . . , hm } is an F -linearly independent subset of L of size m. Since m = q2 , it now follows
from Lemma 2.3 that

  
Nom( M ) = τ : L → F  τ (M ) < F

has size strictly less than qn−1 . Here, of course, each such τ is an F -linear functional.
On the other hand, note that C = {c ∈ L | η(c ) = 1} is a coset of the kernel of η and hence has
size qn−1 . Thus, since 0 ∈
/ C , this set gives rise to precisely qn−1 distinct F -linear functionals τ : L → F
by taking τ (x) = ηc −1 (x) = η(c −1 x) for all c ∈ C . But |Nom( M )| < qn−1 , so there exists d ∈ C with
η(d−1 M ) = F and hence with η(d−1 H ) = F . Of course, d ∈ C says that η(d) = 1.
Now, we know that Ĩ = I ∩ K [ L 2 ] is an L • -stable ideal of K [ L 2 ] that is contained in the maximal
L • -stable ideals corresponding to at least the characters χ̃h,1 with h ∈ H . Thus Ĩ is contained in the
kernel of the algebra homomorphisms corresponding to the characters

−1 x+dy )
χ̃hd−1 ,d (x, y ) = εμη(hd

for all h ∈ H . Since η is an F -linear functional with η(d) = 1, we have

     
η hd−1 x + dy = η hd−1 x + η(d) y = η hd−1 x + y

for all x, y ∈ F . Thus the restriction of χ̃hd−1 ,d to F 2 is given by

−1 )x+ y )
χ̄ (x, y ) = εμ(η(hd .

Since η( Hd−1 ) = F , we see that I ∩ K [ F 2 ] = Ĩ ∩ K [ F 2 ] is an F • -stable ideal of K [ F 2 ] contained

in the kernel ideals associated to all orbits except possibly O0,0 and O∗,0 . But the situation here is
really right–left symmetric, and we know that I ∩ K [ F 2 ] is contained in the ideal corresponding to
O0,∗ , so it must also be contained in the ideal corresponding to O∗,0 . This leaves only ω( F 2 ; F 2 ), the
ideal corresponding to O0,0 . For this, note that 0 ∈ H , so Ĩ is contained in the ideal corresponding to
the orbit χ̃0,t for all 0 = t ∈ L. We can, of course, choose a suitable nonzero element t with η(t ) = 0,
and hence the restriction of this character to F 2 is trivial. We conclude that I ∩ K [ F 2 ] is contained in
all maximal F • -stable ideals of K [ F 2 ], and consequently I ∩ K [ F 2 ] = 0, as required. 2

As an immediate consequence, we have

Lemma 3.5. Let E ⊇ L ⊇ F be finite fields with | F | = q. Suppose that | E : L |  3, | L : F |  2q2 + 1, and let E 2
admit the torus action of E • . If I is an E • -stable ideal of K [ E 2 ] with I ⊆ ω( E 2 ; E 2 ) and I ∩ K [ F 2 ] = 0, then
I = ω( E 2 ; E 2 ) or ω( E ⊕ 0; E 2 ) or ω(0 ⊕ E ; E 2 ) or ω( E ⊕ 0; E 2 ) ∩ ω(0 ⊕ E ; E 2 ).

Proof. We know that I is an intersection of certain maximal E • -stable ideals of K [ E 2 ], and let J be
one of these ideals. If J is standard, then the preceding lemma implies that J ∩ K [ F 2 ] = 0 and hence
I ∩ K [ F 2 ] = 0, a contradiction. Thus J can only be one of the two quasi-augmentation ideals J 1 and
J 2 , and also ω( E 2 ; E 2 ), which we know does occur by assumption. Since J 1 ∩ ω( E 2 ; E 2 ) = ω( E ⊕ 0; E 2 )
and J 2 ∩ ω( E 2 ; E 2 ) = ω(0 ⊕ E ; E 2 ), the result follows. 2
 D.S. Passman / Journal of Algebra 324 (2010) 3035–3043 3043

It is now a simple matter to prove the main result of this paper. We will use the general machinery
developed in [7, Section 1] even though some of this machinery could be fairly easily avoided in the
present context.

Proof of Theorem 1.1. Let F be an infinite locally finite field of characteristic p > 0 and let G = F •
act on V = F 2 as the torus. Then G acts on the group algebra K [ V ] with char K = p, and we begin
by assuming that K contains a primitive pth root of unity. We first show that if I is a nonzero
G-stable ideal of K [ V ] with I ⊆ ω( V ; V ), then I is a finite intersection of the augmentation ideals
corresponding to certain G-stable subgroups of V .
To this end, since I = 0, there exists a finite subfield F 0 of F with I ∩ K [ F 02 ] = 0. If | F 0 | = q, choose
a finite subfield L 0 of F with L 0 ⊇ F 0 and with | L 0 : F 0 |  2q2 + 1. Next let E 0 be a finite subfield of
F with E 0 ⊇ L 0 and | E 0 : L 0 |  3. Since F is countably infinite  and locally finite, we can now find a
chain of finite subfields E 0 ⊆ E 1 ⊆ E 2 ⊆ · · · of F with F = i E i . Notice that {( E 2i , E •i ) | i = 0, 1, 2, . . .}
is a local system for ( V , G ) in the notation of [7, Section 1]. Furthermore, for each i, the preceding
lemma, applied to the fields E i ⊇ L 0 ⊇ F 0 , implies that I ∩ K [ E 2i ] is an intersection of augmentation
ideals of E •i -stable subgroups of E 2i . Thus, by [7, Lemma 1.2], I is an intersection (possibly infinite)
of augmentation ideals corresponding to G-stable subgroups of V . But, by Lemma 1.2, there are only
three nonidentity G-stable subgroups of V , namely V , F ⊕ 0 and 0 ⊕ F , so this intersection must be
finite.
Next, we show that any G-stable ideal of K [ F 2 ] is contained in ω( V ; V ) and hence has the above
form. To this end, observe that Lemma 1.2 implies that all G-sections of V are infinite. Hence by
[7, Lemma 1.3], if A is any G-stable subgroup of V , then ω( A ; V ) is a G-prime ideal of K [ V ]. In
particular, K [ V ] itself is G-prime. Now let I be a G-stable ideal of K [ V ] and assume that I = 0. Then,
by G-primeness, J = I ·ω( V ; V ) is a nonzero G-stable ideal contained in ω( V ; V ). If J = ω( V ; V ), then
I ⊇ J implies that I = ω( V ; V ). Otherwise, by the result of the previous paragraph, J ⊆ ω( A ; V ) for
some G-stable subgroup A properly smaller than V . But ω( A ; V ) is a G-prime ideal and I ·ω( V ; V ) ⊆
ω( A ; V ), so we conclude that I ⊆ ω( A ; V ) ⊆ ω( V ; V ), as required.
We now know that any nonzero G-stable ideal of K [ V ] is contained in ω( V ; V ) and hence is one
of four possibilities, namely ω( V ; V ), ω( V 1 ; V ), ω( V 2 ; V ) and ω( V 1 ; V ) ∩ ω( V 2 ; V ), where V 1 = F ⊕ 0
and V 2 = 0 ⊕ F . In other words, the theorem is proved when K contains a primitive pth root of
unity.
It remains to consider arbitrary fields K with char K = p. In this situation, we let K be an extension
of K that contains a primitive pth root of unity. If I is a nonzero G-stable ideal of K [ V ], then K · I is
a nonzero G-stable ideal of K [ V ], and the freeness of K over K easily implies that ( K · I ) ∩ K [ V ] = I .
Since K · I is an intersection of augmentation ideals ω K ( A ; V ) and since ω K ( A ; V ) ∩ K [ V ] = ω( A ; V ),
the result follows. 2

We remark that the same techniques can be used to handle the case of the torus G of SLn ( F )
acting on F n for any integer n. The result one obtains is then quite analogous to that of Theorem 1.1,
and a full proof will appear elsewhere.

References

[1] C.J.B. Brookes, D.M. Evans, Augmentation modules for affine groups, Math. Proc. Cambridge Philos. Soc. 130 (2001) 287–294.
[2] R.A. Brualdi, Introductory Combinatorics, fifth ed., Pearson, Upper Saddle River, NJ, 2010.
[3] D. Goldstein, R. Guralnick, L. Small, E. Zelmanov, Inversion invariant additive subgroups of division rings, Pacific J. Math. 227
(2006) 287–294.
[4] J.M. Osterburg, D.S. Passman, A.E. Zalesskiı̆, Invariant ideals of abelian group algebras under the multiplicative action of a
field, II, Proc. Amer. Math. Soc. 130 (2002) 951–957.
[5] D.S. Passman, The Algebraic Structure of Group Rings, Wiley–Interscience, New York, 1977.
[6] D.S. Passman, Invariant ideals of abelian group algebras under the action of simple linear groups, Resenhas IME-USP 5 (2002)
377–390.
[7] D.S. Passman, A.E. Zalesskiı̆, Invariant ideals of abelian group algebras under the multiplicative action of a field, I, Proc. Amer.
Math. Soc. 130 (2002) 939–949.